PERSONALITY PROCESSES AND
DIFFERENCES
Traits, Consistency, and Self-Schemata:
What Do Our Methods Measure?
Pamela A. Burke, Robert E. Kraut, and Robert H. Dworkin
Cornell University
Psychologists have responded to the inadequacies that Mischel (1968) noted in
the trait approach to personality by exploring two other facets of personality,
cross-situational consistency and self-schemata. These newer approaches have yet
to be clearly distinguished conceptually or empirically from the traditional model
that they were designed to supplement or replace. Our research tries to do this.
In it, two samples of respondents (N = 362) rated the extent to which, of 10
traits applied to them (overall level), rated their consistency on these traits (crosssituational consistency), and rated the importance of these traits to their view of
themselves (self-schema). Correlational analyses show that the measures of consistency and self-schema lack discriminant validity from the measures of overall
level. Specifically, their correlations with level were as high as their internal
consistencies. We argue that the measurement models for cross-situational consistency and for self-schemata do not adequately reflect their theoretical counterparts.
This failure undercuts the interpretations of recent research by Markus (1977),
Markus, Crane, Bernstein, and Siladi (1982), and S. Bern (1981).
To a great extent the results of psychologists' attempt to understand personality has
been a list of personality traits and research
on their causes, correlates, and consequences.
But Mischel (1968) and others (Peterson,
1968; Vernon, 1964) severely questioned the
value of this enterprise. One of the responses
to Mischel's criticism of the trait approach
to personality has been the introduction of
two alternative ways of thinking about personality—the cross-situational consistency
approach and the self-schema approach. Our
intention in this article is to examine the
conceptual distinctions among the facets of
personality posited by these three approaches
and the methods for measuring them. We
Portions of this research were supported by National
Science Foundauon Grant BNS-7907963 to the second
author. An earlier version of this research was presented
at the Eastern Psychological Association Meeting, New
York, 1981. Robert Kraut is now at Bell Communications
Research, Murray Hill, New Jersey. Robert Dworkin is
now at the Departments of Anestheaology and Psychiatry,
College of Physicians and Surgeons. Columbia University,
New York, New York.
Requests for reprints should be sent to Pamela Burke,
who is now at the Department of Psychology, Building
420, Jordan Hall, Stanford University, Stanford, California
94305.
especially concentrate on the self-schema approach.
Overall Level
The traditional trait approach has generally
regarded a trait (the extent to which an
individual is high, moderate, or low on a
personality dimension) as a latent tendency
to act consistently with one's level on that
dimension (Magnusson & Endler, 1977). We
refer to this measurement as overall level. As
Magnusson and Endler note, many of the
arguments about the usefulness of this way
of conceptualizing personality come from
empirical refutations of the trait measurement
model and not from theoretical disagreements.
The trait measurement model makes two
assumptions that lead to predictions unsubstantiated by empirical research. The model
assumes that each individual possesses a true
score on each personality trait and that an
accurate measurement of this score will be
monotonically related to each behavioral
measure of that trait. These assumptions
imply that stable rank orders of individuals
exist across situations in the expression of
trait-related behavior.
Journal of Personality and Soaal Psychology. 1984. Vol 47, No 3, 568-579
Copyright 1984 by the American Psychological Association. Inc
568
TRAITS, CONSISTENCY, AND SELF-SCHEMATA
569
A person's overall level on a trait is generally measured through the use of self-reports. Single item questions that ask directly
for the person's extremity on the trait and
personality inventories, (i.e., multiple item
tests that produce a composite score on a
trait, e.g., California Personality Inventory,
[CPI] Gough, 1969), are typically used.
Theoretical and empirical justification for the
use of single item global measures of traits
(Kenrick & Braver, 1982; Peterson, 1965)
continues alongside research demonstrating
the efficacy of composite scales in behavioral
assessment (Epstein, 1979; Moskowitz, 1982;
Rushton, Jackson, & Paunonen, 1981).
model. Schema theorists think about traits in
terms of cognitive generalizations about the
self. Like consistency theorists, they dismiss
the assumption of a one-to-one correspondence between a person's level on a trait and
his or her trait relevant behavior. People's
self-schemata are not dependent on their
overall level on a trait but rather on the
personal significance or salience of the trait
dimension to them. That is to say, their
cognitive organizations determine which dimensions are salient and important to them.
According to this view, people with different
schemata interpret and respond to the same
stimuli in different ways.
Cross-Situational Consistency
Systematic effects in social behavior depend less on
people having some amount of a particular substantive
attribute, such as independence or dependence, and more
on the readiness or ability to categorize behavior along
certain dimensions. (Markus, 1977, p. 77)
Cross-situational consistency theorists have
responded to the apparent unpredictability
of behavior demonstrated by the traditional
trait measurement model by specifying the
conditions under which traits predict behavior.
They assert that behavioral consistency is a
function of both the meaning of the trait and
the meaningfulness of that trait to the individual. This approach maintains that stable
cross-situational consistency in behavior is
characteristic of people for some, but not all,
traits. Thus, the accurate prediction of behavior depends on knowledge of the traits on
which an individual is consistent and knowledge of the behaviors he or she sees as
relevant to those traits. As a result, according
to these theorists, traditional trait research
that measures only a person's level on a trait,
has underreported the predictability of behavior.
Consistency researchers have attempted to
demonstrate that measuring self-reported
consistency improves the prediction of behavior. For example, D. Bern and Allen (1974)
have shown that, at least for the trait of
friendliness, people can report how consistent
they are across situations. They argue that
one can predict friendly behaviors from measures of overall level better for consistent
people than for inconsistent ones (cf. Mischel
& Peake, 1982).
The Measurement Problem
We have sketched the conceptual relations
between the traditional level model of traits
and two alternative models, a cross-situational
consistency model and a self-schema model.
They both hold promise of remedying the
inability of the traditional trait model to
strongly predict behavior. Before they are
accepted, however, their empirical relations
to the traditional measurement model of
traits must be examined. If they show only a
small relation to overall level, then a combination of the new models with the traditional
model might predict behavior better than
either alone. On the other hand, despite their
conceptual independence, the new measures
of traits may be highly related to the traditional level measure. In this case little is
gained by hiding the use of traditional methods under new labels.
Previous research leaves the relation among
the three models of personality in doubt. The
standard method of measuring cross-situational consistency in personality precludes
assessing its natural relation with overall level.
For example, D. Bern and Allen (1974) and
Zanna, Olson, & Fazio (1980) statistically
controlled level when measuring consistency.
Self-Schema
Stones and Burt (1978) and Rushton et al.
A theory of self-schema is another response (1981), however, have found that the more
to Mischel's critique of the traditional trait. extremely people rate their level on a personal
570
P. BURKE, R. KRAUT, AND R. DWORKIN
trait, the more they also rate themselves as Measures
consistent on that trait For example, people
The questions were self-report items designed to assess
who rate themselves as moderate on a trait participants' trait level, trait consistency, and self-schema
also rate themselves as highly variable across for the three traits of friendliness, intelligence, and articsituations. Other measures of cross-situational ulateness. We used friendliness and intelligence as target
consistency (Kenrick & Stringneld, 1980) have traits because of their repeated assessment in the personality literature and because of their independence. We
been criticized for ignoring trait extremity used articulateness because it is a more specific trait with
entirely (Rushton et al., 1981).
clear behavioral manifestations. Questions about other
Self-schema theorists have confounded traits were added as filler items so that the target traits
would not be recognized as the focus of inquiry. The
measures of schema and measures of level specific
questions used are shown in Table 1. The reliand, therefore, have obscured the relation abilities and validities of several of these measures are
between peoples' schemata about a trait and unknown. They were chosen because they are the meatheir level on that trait. Thus for researchers sures commonly used by researchers in the field or are
such as Markus (Markus, 1977; Markus et directly derivative of the proposed theories.
Level. The four measures were traditional self-report
al., 1982) and S. Bern (1981; 1982), a person liken
scales, on which participants reported the extent
who is schematic on a trait is one who has to which a trait described them, either in general, coman extreme level on that dimension and who pared to their peers, or in particular social circumstances,
considers the dimension important. The re- (e.g., In general, how [friendly] are you?)
Self-schema. Schema theorists like Markus (1977)
lation between level and importance is unand S. Bern (1981) define self-schema as the importance
clear.
of and readiness to use a trait dimension in thinking
Finally, the relation between consistency about one's self and the world. However, they measure it
and self-schema is also of interest. For ex- by combining measures of importance with measures of
ample, although Markus (1977) posits a pos- level. To determine the relation of the importance of a
for a person to that person's level on the
itive relation between the two, no data yet dimension
dimension, we isolated these components. We measured
exist to support this claim.
self-schema by asking questions about the importance
Overview
The empirical goal of the present article is
to test the discriminant validity of methods
to measure peoples1 overall level on a trait,
their cross-situational consistency on that
trait, and their self-schema on that trait We
used several self-report measures of trait level,
cross-situational consistency, and self-schema
that were obtained from previous research or
designed for the present studies. Participants
described themselves on these measures for
each of several traits. The correlations between
all measures were computed to assess the
naturally occuring relations among level,
consistency, and schema and to evaluate the
convergent and discriminant validity of these
facets of personality.
Study 1
Method
Participants and Procedures
One hundred and ninety male and female college
undergraduates at Cornell University were paid $.50 to
complete a brief questionnaire.
people assigned to a trait (e.g.. How important is friendliness . . . to the image you have of yourself, regardless
of whether or not the trait describes you? [from Markus,
1977]. We also asked about their readiness to use the
trait in thinking about themselves and others, (e.g., When
attending a lecture, how frequently do you judge how
friendly the speaker is?)
Cross-sittuuional consistency The questions here were
modeled after those used by D. Bern and Allen (1974).
They included both participants' self-reports on consistency (e.g., How much do you vary from one situation
to another in how friendly you are?) and the standard
deviation of their self-reported level in different social
circumstances (e.g., friendliness with parents, friends, or
siblings).
Results
The central datum of this article is a
comparison of the internal consistency of the
various level, schema, and consistency measures with their discriminant validity. We
computed Pearson correlations between all
measures of level, schema, and consistency
separately for the traits of intelligence, friendliness, and articulateness. The signs and magnitudes of the correlations were highly similar
across the three traits. We have presented the
comparison twice, once averaging over measures and once averaging over traits.
TRAITS, CONSISTENCY, AND SELF-SCHEMATA
571
Table 2 shows the z-transformed correla- item correlation matrices of the three traits
tions for the three traits of friendliness, intel- to test the independence of the three facets
ligence, and articulateness, averaged over of personality. We performed a principal axis
measures. Table 1 presents the mean z-trans- factor analysis with oblique rotation separately
formed correlations for the different measures for each trait The hypothesis of independence
of level, consistency, and schema averaged would predict a three factor solution with
over traits.
measures of level, self-schema, and consisAs Tables 1 and 2 show, the internal con- tency loading primarily on three distinct facsistency of the traditional trait level measures tors. Instead, at least four factors were ob(mean interitem correlations = .34) was at tained for each trait (eigenvalue > 1.0) and
least as high as that of the schema measures more important, measures of level, self(mean interitem correlations = .20) or the schema, and consistency do not cohere on
cross-situational consistency measures (mean separate factors. The factors, though highly
consistent across traits, show a complex patinteritem correlations = .23).'
Tables 1 and 2 also show the cross-corre- tern of relations among the measures, the
lations between the various models of person- specific nature of which could not have been
ality. The discriminant validity of both the predicted in advance.
schema and the cross-situational consistency
Table 3 presents the results from the analmodels of personality was weak. Specifically, ysis of the measures of friendliness. Four
the mean correlation between measures of factors account for 63% of the variance in
schema and measures of level (mean r = .23) these measures. The major factor (Factor 1)
was as high as the internal consistency of the for friendliness and for the other two traits
schema measures (mean r = .20). This means consists of two level measures and three selfthat items designed to measure schema cor- schema measures (In Table 3, items 1, 2, 5,
relate as highly with items designed to mea- 6, and 7). The schema measures involving
sure level as they do with each other. The the judgments of others also load on this
two consistency measures show no linear factor for two traits (items 10, 11, and 12).
relation to either the level measures (mean For all traits, Factor 2 is composed of level
r = .14) or to the schema measures (mean judgments on a trait when the judge is with
r = .04).
significant others and the self-schema measure
Unlike Stones and Burt's (1978) data, we of confidence (items 3 and 8). In addition,
found no curvilinear relation between consis- across the three analyses, other level and selftency and trait level. We computed polariza- schema measures load inconsistently on this
tion scores for each participant on each trait, factor. Factor 3 for all three traits is composed
by taking the absolute value of the difference of the consistency measures. It also includes
between each participant's score on a measure a mixed pattern of both level and self-schema
and the group mean on that measure. The measures. Factor 4 consists of the level meacorrelations of the consistency measures with sure comparing self to peers, and an inconthe polarization scores are low (mean sistent pattern of other level, self-schema, and
consistency measures.
T = .06).
The argument against the internal consisStudy 2
tency and discriminant validity of the selfschema and the cross-situational consistency
Method
measures is based on the patterns of correlations we have presented and not on their Participants and Procedures
One hundred and seventy-two undergraduate women
absolute magnitudes. Given the way that we
aggregated correlations over measures and from Cornell University were recruited through newspaper
traits and the nonindependence of the correlations, we could not directly test the signifi1
The reliability for the cross-situational consistency
cance of the difference between mean corre- measure is very similar to the interitem correlations that
lations. Therefore, to examine these patterns D. Bern and Allen (1974) found for similar
more explicitly, we factor analyzed the inter- (r •= .22).
Table 1
Correlations Among All Measures of Level, Self-Schemata. and Consistency, Averaged Across Three Trail Dimensions: Study 1
Self-schemata
Level
Measures
Level
1. In general, how
are you?
2. Trait
describes
me/does not describe
me.
3. How
are you
with your parents,
friends, etc.? (Mean of
10 judgments)
4. Compared to your peers,
how
are you?
Self-schemata
5. How important is trait
in describing
yourself honestly to
someone?
6. Importance of trait
to your image
of yourself?
7. Importance of trait
in
understanding your
personality and
behavior?
8. Confidence ratings of how
you are with
10 specific others.
9. How easy is it for you to
judge yourself on the
trait
?
10. How easy is it to judge a
friend on trait
?
11. Importance of trait
in
understanding a
friend's personality and
behavior?
t
Consistency
2
3
4
5
6
7
8
9
10
11
12
13
14
.73
.44
.14
.38
.40
.46
.19
.16
.20
.22
.23
.24
.06
.47
.12
.37
48
.37
.19
.19
.24
.25
.29
.13
.19
.14
.28
.27
.25
.48
.19
.18
.11
.20
.17
.23
.07
.16
.09
.10
.09
.15
.12
.12
.03
-.01
.39
.40
.16
.15
.24
.37
.25
.13
.01
51
.11
.10
.18
.31
.28
.09
.12
1
.13
.08
.14
.31
.29
.11
.07
.09
.09
.06
.08
.12
-.09
.06
.03
.14
.06
-.01
.26
.14
.01
.04
.21
.05
.05
TRAITS, CONSISTENCY, AND SELF-SCHEMATA
o
573
advertisements and posters. They completed a set of
personality questionnaires and were paid $2.50.
3
'in
a
1
Measures
8
The questionnaires we used in this study were originally
developed to measure the three facets of personality for
the trait of friendliness, as part of a study on memory
for events. In addition, we included a less complete set
of measures of the three facets for the traits of conscientiousness, independence, honesty, sensitivity, anxiety, activity level, and assertiveness
Trait level. At least one of three different measures
of level were obtained for each trait dimension. These
included global self-reports for all eight trait dimensions
(e.g., In general, how friendly are you?), self-reports about
five traits in a number of different situations (The CrossSituational Behavioral Survey (CSBS), D. Bern & Allen,
1974), and a reliable personality inventory (CPI Sociability
Scale).
Self-schema We adapted two versions of Marfcus's
(1977) self-report measures of self-schema to make them
conceptually independent of overall level. Each was administered to approximately half the sample. In addition,
we adapted the trait-cued autobiography method used
by McGuire and Padawer-Singer (1976). Participants
were asked to spend IS mm writing a short description
of those aspects of their personality that they thought
were most important in understanding their own behavior.
They were given a list of 24 traits they could choose to
use, "because previous participants have found the list
helpful in organizing their answers". Two coders coded
these short essays as schematic for a trait dimension if
the participant used a relevant cue from the cue list, a
synonym of it, or an antonym for it. When the coders
disagreed on the presence of a cue, the participant was
classified as schematic because at least one of the coders
had found some mention of the dimension in the autobiography. We obtained the trait-cued autobiography
measure of self-schema for seven traits.
Cross-situational consistency. We measured crosssituational consistency for each trait through at least one
of the two measures used by D. Bern and Allen (1974).
These included a direct self-report (e.g., "How much do
you vary from one situation to another in how friendly
you are?") for all eight traits and the variance in participants' level judgments in a number of social situations
(CSBS) for five traits.
Results
|
Pearson product-moment correlations were
computed among the measures of level, crosssituational consistency, and self-schema. Table
4 presents means of the correlations for the
eight trait dimensions averaging across measures for each trait, and Table 5 presents the
means of the correlations averaging across
traits for each measure. As in Study 1, measures of overall level show the most convergence, with a mean correlation across measures and traits of .57. The mean internal
consistency of the schema measures (r = .21)
574
P. BURKE, R. KRAUT, AND R. DWQRKIN
Table 2
Correlations Between Measures of Level,
Self-Schemata, and Consistency for the
Traits of Friendliness. Intelligence, and
Articulateness: Study 1
Schema As a Processing Model
Markus and Bern might dismiss this criticism of the schema model. Markus (1977)
and S. Bern (1981) denned self-schema in
terms of the personal significance of wellConsistency
Level
Schemata
Measure
articulated cognitive structures, yet they originally and intentionally measured these strucFriendliness
tures in terms of overall level. Markus and
.17
.32
.22
Level
her colleaques (1982) have dropped personal
.09
.18
Schemata
significance or importance from their defini.24
Consistency
tion of self-schema and yet still consider the
Intelligence
schema model a supplement to or an alternative to the traditional trait model of perLevel
.37
.28
.13
Schemata
.22
.06
sonality. They would argue that the schema
Consistency
.24
model is valuable, even if one cannot measure
it independently of a person's level on a trait.
Articulateness
The schema model provides the cognitive
Level
.14
.33
.20
mechanism through which trait level has its
Schemata
.20
.04
effects
on beliefs and behavior.
Consistency
.22
The mechanism is the workings of an
autonomous belief system, formed through
the differential experience of people who differ
and of the cross-situational consistency mea- on their level on a trait.
sures (r = .22) were substantially less.
Tables 4 and S also show little discriminant With social experience we gain a diversity of self-relevant
information that becomes organized into cognitive strucvalidity among the various measures. As in tures. It is by means of these structures that we categorize,
Study 1, the mean correlations of level and explain, and evaluate our behavior in various focal
schema (r = .30) and of level and cross-situ- domains, (Markus et al., 1982, p. 38).
ational consistency (r = .34) were as high as
the internal consistency of the self-schema Markus and Bern would argue that the schema
concept is the preferred explanation for the
and consistency measures.
individual differences in the way people process information about themselves. According
to this reasoning, the value of the schema
Summary and Discussion
model rests on the discriminant validity of
We have shown that measures of self- underlying cognitive processes and not on the
schema and cross-situational consistency lack discriminant validity of a personality meadiscriminant validity. Table 6 summarizes surement.
For example, Markus and her students
our evidence by averaging the convergent
validity correlations and the discriminant va- have found that compared with people who
lidity correlations across traits, measures, and are aschematic, people who score as selfstudies. It shows that measures of self-schema schematic on a trait dimension make schemaand cross-situational consistency, although related judgments about themselves more
designed to be independent of overall trait quickly and consistently, offer more behavioral
level, are not independent. Instead the cor- evidence for their self-judgments, make more
relation of each with overall level is as high confident and extreme predictions about fuas its internal consistency. The measurement ture behavior, and change their self-concepmodels for the cross-situational consistency tions less when given counterevidence.
approach and the self-schema approach to
Both Markus and S. Bern conclude from
personality clearly Ml to reflect the conceptual this type of evidence that schematic people
models.
have an elaborated cognitive structure that
TRAITS, CONSISTENCY, AND SELF-SCHEMATA
enables them to process relevant information
about themselves quickly and efficiently. They
both assume that the self-inference strategies
that schematics use are qualitatively different
from those used by aschematics. Having assumed the existence of schemata, researchers
are now trying to distinguish between theoretical types, particularly between self-schemata and gender schemata. (See S. Bern,
1982, for a skillful explication of differences
between the two.)
However, we believe this investigation is
premature. The data on the speed of, conndence in, and richness of self-descriptions are
not compelling evidence that some people
form their self-descriptions through any sort
of schema processing. We have shown that
putative schematics and aschematic individuals also differ on trait level. We believe their
different levels on traits can account for the
575
data, without postulating qualitatively different processing strategies between schematic
and aschematic individuals,
For example, consider the data on speed
of processing. Both Markus (Markus, 1977;
Markus et al., 1982) and S. Bern (1981; citing
data from Girvin, 1978 and Markus et al.,
1982) argue that the greater speed with which
schematic compared with aschematic participants endorse schema-consistent personality
descriptions is evidence for the schema model
of personality. According to these authors, to
answer the question, "Are you independent?"
or "Are you nurturant?" schematic people
need only look up the correct answer in their
preexisting and well-structured self-descriptions or gender-schemata, whereas aschematic
people must use different cognitive processes,
for example, reviewing and evaluating dimcult-to-abstract behavioral evidence. However,
Table 3
Principle Component Analysis With Oblique Rotation for Measures of Friendliness
Measures
Level
1. In general, how friendly are you?
2. Friendly described me/does not describe me.
3 How friendly are you with your parents, friends,
etc.? (Mean of 10 judgments)
4. Compared to your peers, how friendly are you?
Self-schemata
5. How important is friendliness in describing yourself
honestly to someone?
6. Importance of friendliness to your image of
yourself?
7. Importance of friendliness in understanding your
personality and behavior?
8. Confidence ratings of how friendly you are with 10
specific others.
9. How easy is it for you to judge yourself on
friendliness?
10. How easy is it to judge a friend on the trait
friendly?
11. Importance of friendliness in understanding a
friend's personality and behavior?
12. When attending a lecture, how frequently do you
judge the speaker on friendliness?
Consistency
13. How much do you vary from one situation to
another in how friendly you are?
14. Standard deviation of how friendly you are with 10
specific others.
Factor
1
Factor
2
Factor
3
Factor
4
.73
.81
.41
.44
-.15
-.02
.14
.18
.38
.18
.84
.24
-.09
-.23
.11
-.64
.68
.16
.16
.09
.77
.05
-.21
.29
.74
-.03
-.09
.28
.15
.77
-.25
.35
.03
.40
.16
-.04
.10
.13
-.03
.76
.41
-.23
.31
.64
.43
.16
.64
.19
-.38
-.01
.56
-.40
-.22
-.19
.63
-.08
576
P. BURKE, R. KRAUT, AND R. DWORKJN
we can adduce at least two nonschema explanations for why schematic people respond
faster than aschematic ones.
The threshold account. A threshold account assumes that people apply a trait dimension to themselves when their level on
that trait exceeds some subjective threshold
and that, on the average, this threshold does
not systematically differ among people. Quick
responses occur when the individuals' level
on the trait is unambiguously distinct from
the threshold, that is, when they are extreme
on the trait. A closer, more time consuming
comparison is required for traits near the
subjective threshold, that is, for traits judged
moderately self-descriptive. Threshold for-
Table 4
Correlations Between Measures of Level. SelfSchemata, and Consistency Presented by TraitStudy 2
Measures
Friendliness
Level
Schemata
Consistency
Sensitivity
Level
Schemata
Consistency
Independence
Level
Schemata
Consistency
Honesty
Level
Schemata
Consistency
Conscientiousness
Level
Schemata
Consistency
Assertiveness
Level
Schemata
Consistency
Activity Level
Level
Schemata
Consistency
Anxiety
Level
Schemata
Consistency
Level
Schemata
.62
.22
.20
.40
.24
.27
.49
.31
.33
.41
.09
.19
t
.26
.19
.53
.19
Consistency
•
.45
.30
.27
.37
.09
.13
.69
.33
.23
.49
.19
.30
.59
.20
•
.20
.00
.20
.33
.15
.42
.16
.42
.16
-.43
-.31
a
" Only one measure of this facet was collected for this
trait.
mutations of the self-referential judgment
process have garnered considerable empirical
support (Goldberg, 1963; Jackson & Helmes,
1979; Jackson & Paunonen, 1980; Rogers,
1971; Rogers, Kuiper, & Rogers, 1979; Voyce
& Jackson, 1977) in the personality measurement literature. Consistent with this formulation, consistency and latency of response to
self-ascriptions for a given dimension have
been shown to depend on the individual's
level on that dimension (Kuncel, 1973). According to the threshold account, aschematics
take longer because of the difficulty they have
in making discriminations when their level
on a trait is close to their subjective threshold
for calling it self-descriptive.
A similar argument can be made to explain
findings (Girvin, 1978; Markus et al., 1982)
that sex-typed (i.e., schematic) people were
faster at judging sex-congruent traits to be
self-descriptive and slower at judging sexincongruent traits to be self-descriptive. Nonsex-typed (i.e., aschematic) people did not
show these differences in speed of response.
By definition, the schematic people are more
extreme in the positive direction on sexcongruent traits judged self-descriptive than
on cross-sex-congruent traits judged self-descriptive. On the average, a schematic man
would judge himself as more assertive than
he is artistic, even though he might apply
both terms to himself; any deviations would
presumably be the result of measurement
error. Aschematic people are by definition
equally extreme on sex-congruent and crosssex-congruent traits judged self-descriptive.
For example, an aschematic man might judge
himself to be highly assertive and artistic,
and equally so. As a result, the latency differences between the two types of traits observed
for schematic but not for aschematic participants reflects aschematic individuals' choice
between equally applicable traits.
Evaluation of behavioral evidence. Evaluating behavioral evidence is a second process
through which people can decide whether a
trait is self-descriptive. This process can also
account for the differences between schematic
and aschematic individuals' speed of response.
Those who show an extreme level on a trait
are likely to have had mare and more consistent occasions in which they have acted consistently with the trait In reviewing evidence
577
TRAITS, CONSISTENCY, AND SELF-SCHEMATA
TableS
Correlations Among all Measures of Level. Self-Schemata, and Consistency Averaged
Across Eight Trait Dimensions: Study 2
Level
Self-schemata
Consistency
Trait dimensions
1. California Personality
Inventory
Sociability item
2. In general, how
are you?
3. Cross-Situational
Behavior Survey
Scale mean
4. Trait-cued
autobiography
5 Importance selfreport I
6. Importance selfreport II
7. How much do you
vary from one
situation to
another in how
you are?
8. Cross-Situational
Behavior Survey
Scale variance
.53
.64
.03
.33
.08
.37
.32
.59
.20
.52
.36
.36
.18
.14
.29
.35
.43
.44
.22
.20
.03
.01
.21
.11
.15
.03
.22
about their independence, for example, schematic people will be able to retrieve occasions
during which they acted independently more
quickly. In addition, the evidence they retrieve
should be more internally consistent. As a
result, they should have an easier and quicker
time integrating the evidence and then making
their decision.
Similar reasoning can account for Markus'
behavioral instance data and subjective likelihood data. These data show that people
offer the greatest behavioral evidence for those
traits that are most self-descriptive and view
engaging in behaviors related to these traits
in the future as relatively likely.
Table 6
Correlations Among Measures of Level. Selfschemata, and Consistency Averaged Across
Measures and Traits for Studies 1 and 2
Measures
Level
Schemata
Consistency
Level
Schemata
Consistency
.41
.26
.20
.26
.08
.22
Cross-Situational Consistency Evidence
Our data on cross-situational consistency
measures appear somewhat at odds with the
boost in predictive validity reported by researchers who have used them in the past.
We would not argue on the basis of our
sparse data on these measures that trait level
and consistency as commonly measured are
completely overlapping constructs. The utility
of using consistency and schema measures as
moderator variables after trait level has been
statistically controlled remains an empirical
question. Our data are consistent, however,
with recent research by Mischel and Peake
(1982), which strongly questions the predictive
utility of the trait based cross-situational consistency approach. Mischel and Peake (1982)
report that both in D. Bern and Allen's
(1974) original study and in their own replication of that study that people who report
high self consistency on a trait do not show
greater cross-situational consistency in behavior than those people who report low
consistency. (They do, however, evoke greater
consistency in other people's descriptions of
578
P. BURKE, R. KRAUT, AND R. DWORKIN
their behavior.) Consistency models empirically related to traditional trait models have
not been proven to provide any great enhancement in behavioral predictability beyond a simple trait level model.
and as measured will enable personality prospectors to extract the different insights of
each or persuade them to seek, ultimately,
an entirely new stream in which to pan.
References
Conclusions
In the aftermath of Mischel's critique of
the trait approach to personality, personality
theorists adopted two responses. One was to
emphasize an interactionist approach to personality (e.g., Magnusson & Endler, 1977),
and the other was to explore other models of
personality; the cross-situational consistency
model and the self-schema model. We have
shown that at least operationally, neither of
these models is distinguishable from the traditional trait model it was designed to supplement or replace. We have shown that
measures of self-schema and of cross-situational consistency correlate as highly with
overall trait level as they do among themselves.
In addition, we have argued that the concept
of trait level (and the different experiences
that go with different trait levels) can account
for the evidence about individual differences
in the speed, confidence, and richness of selfdescriptions, without appeal to schema. And,
similarly, we argued that trait level may account for whatever predictive utility trait
based cross-situational consistency models can
claim.
As yet, neither self-schema models nor
cross-situational models have led to a strong
predictive theory of personality. Our data
suggest that these models are not, in practice,
as innovative as once hoped. We concur with
Mischel and Peake who state
the pursuit of person-situation interaction will require a
theoretical reconceptualization of both personality and
situation constructs themselves, not just more clever
methods for applying everyday trait terms to people's
behavior in particular contexts. (1982, p. 748)
As self-schema theorists take their cognitive
process model from the laboratory and attempt to form a predictive personality model
in the social world, they too may find that
dependence on the trait model has undermined their endeavor from the beginning.
Only a thorough evaluation of the relations
between schema, cross-situational consistency,
and traditional trait models both in theory
Bern, D., & Allen, A. (1974). On predicting some of the
people some of the time: The search for cross-situational
consistencies in behavior. Psychological Review, 81.
506-520.
Bern, S. (1981). Gender schema theory: A cognitive
account of sex-typing. Psychological Review. 88, 354364.
Bern, S. (1982). Gender-schema theory and self-schema
theory compared: A comment on Markus, Crane,
Bernstein, and Siladis, "Self-schema and gender." Journal of Personality and Social Psychology. 43. 11921194.
Epstein, S. (1979). The stability of behavior I. On
predicting most of the people much of the time.
Journal of Personality and Social Psychology, 37, 10971126.
Girvin, B. (1978). The nature of being schematic: Sexrole, self-schemas and differential processing of masculine andfeminine information. Unpublished doctoral
dissertation, Stanford University, Stanford, CA.
Goldberg, L. R. (1963). A model of item ambiguity in
personality assessment Educational and Psychological
Measurement, 23, 467-492.
Gough, H. G. (1969). Manual for the California Personality Inventory (rev. ed.). Palo Alto, CA: Consulting
Psychologists Press.
Jackson, D. N., & Helmes, E. (1979). Personality structure
and the circumplex. Journal of Personality and Social
Psychology. 37, 2278-2285.
Jackson, D. N., & Paunonen, S. V. (1980). Personality
structure and assessment Annual Review of Psychology.
31. 503-551.
Kenrick, D. T., & Braver, S. L. (1982). Personality:
Idiographic and nomothetic! A rejoinder. Psychological
Review, 89, 182-186.
Kenrick, D. T, & Stringfield, D. O. (1980). Personality
traits and the eye of the beholder Crossing some
traditional philosophical boundaries in the search for
consistency in all of the people. Psychological Review,
27, 88-104.
Kuncel, R. B. (1973). Response processes and relative
location of subject and item. Educational and Psychological Measurement, 33, 545-563.
Magnusson, D., & Endler, N. (1977). Interactional psychology: Present status and future prospects. In D.
Magnusson & N. Endler (Eds.), (pp. 3-36). Personality
at the Crossroads: Current issues in interactional psychology. Hillsdale, NJ: Eribaum.
Markus, H. (1977). Self-schemata and processing of
information about the self. Journal of Personality and
Social Psychology, 35, 63-78.
Markus, H., Crane, M., Bernstein, S., & Siladi, M.
(1982). Self-schemas and gender. Journal of Personality
and Social Psychology, 42, 38-50.
McGuire, W., & Padawer-Singer, A. (1976). Trait salience
in the spontaneous self-concept. Journal of Personality
and Social Psychology, 33. 743-754.
TRAITS, CONSISTENCY. AND SELF-SCHEMATA
Mischd, W. (1968). Personality and Assessment. New
York: WUey.
Mischel, W., & Peake, P. R. (1982). Beyond deja vu in
the search for cross-shuational consistency. Psychological Review, 89, 730-755.
Moskowitz, D. S. (1982). Coherence and cross-situational
generality: A new analysis of old problems. Journal of
Personality and Social Psychology. 43. 754-768.
Peterson, D. R. (1965). Scope and generality of verbally
defined personality factors. Psychological Review. 72.
48-89.
Peterson, D. R. (1968). The clinical study of social
behavior. New York: Appleton-Century-Crofts.
Rogers, T. B. (1971). The process of responding to
personality items: Some issues, a theory, and some
research. Multivariate Behavior Research Monograph.
6.
Rogers, T. B., Kuiper, N. A., & Rogers, P. J. (1979).
Symbolic distance and congniity effects for pairedcomparisons judgments of degree of self reference.
Journal of Research in Personality, 13, 433-449.
579
Rushton, J. P., Jackson, p . N., & Paunonea, & V. (1981).
Personality: Nomothctic or idiographic? A response to
Kenrick and Stringfidd. Psychological Review, 88,
582-589.
Stones, M., & Burt, G. (1978). Quasi-statistical inference
in rating behavior. 1. Preliminary investigations. Journal
of Research in Personality. 13. 381-389.
Vernon, P. E. (1964). Personality assessment: A critical
survey New York: WUey.
Voyce, C. D., & Jackson, D. N. (1977). An evaluation of
a threshold theory for personality assessment Educational and Psychological Measuement, 37. 383-408.
Zanna, M. P., Olson, J. M.. & Fazio, R. H. (1980).
Attitude-behavior consistency: An individual difference
perspective. Journal of Personality and Social Psychology, 38, 432-440.
Received January 4, 1983
Revision received May 6, 1983